University of Connecticut
DigitalCommons@UConn
University Scholar Projects
University Scholar Program
Spring 5-1-2016
Stability Of Norwalk Virus Capsid Protein
Interfaces Evaluated By In Silico Nanoindentation
Prakhar Bansal
University of Connecticut, [email protected]
Follow this and additional works at: http://digitalcommons.uconn.edu/usp_projects
Part of the Biophysics Commons, and the Structural Biology Commons
Recommended Citation
Bansal, Prakhar, "Stability Of Norwalk Virus Capsid Protein Interfaces Evaluated By In Silico Nanoindentation" (2016). University
Scholar Projects. 22.
http://digitalcommons.uconn.edu/usp_projects/22
STABILITY OF NORWALK
VIRUS CAPSID PROTEIN
INTERFACES EVALUATED
BY IN SILICO
NANOINDENTATION
Author: Prakhar Bansal
Research Advisor: Eric May
Honors Advisor: Eric May
University Scholar Committee: Eric May (chair), Victoria
Robinson, Carolyn Teschke
University of Connecticut
Department of Molecular and Cell Biology
Abstract
Studying the mechanical properties of viral capsids can give several insights into not only the
lifecycle of the virus, but also into potential drug targets to thwart the progression of viral infection.
Nanoindentation using an atomic force microscope is a useful technique for determining structural
properties of small molecules and particles, and is commonly used to study viral capsids. This
technique utilizes the probe of the microscope to push down on the capsid and record the forces
along the indentation path. We ran this experiment in silico where we simulated the nanoindentation
of Norwalk virus capsids using molecular dynamics. Running a simulation of the nanoindentation
allowed us to observe the capsid deformation in much more detail than is possible experimentally.
We were able to identify a distinctly weak interface in the Norovirus capsid. This interface might be
the initial interface to disassemble during viral uncoating in the host cell for infection to proceed.
Strengthening this interface might prevent uncoating and further infection. We identified three sites
in this weak interface that may prove to be good drug targets for an antiviral treatment of Norwalk
virus. Our study culminated in a publication in the journal Frontiers in Bioengineering and
Biotechnology (Boyd, K. J., Bansal, P., Feng, J., & May, E. R. (2015). Stability of Norwalk Virus
Capsid Protein Interfaces Evaluated by in Silico Nanoindentation. Frontiers in Bioengineering and
Biotechnology, 3(July), 1–8. doi:10.3389/fbioe.2015.00103).
Introduction
Norwalk virus (genus Norovirus) is the leading cause of
gastroenteritis worldwide1. Norovirus outbreaks commonly
occur in settings with high levels of contact and less-thanideal hygiene such as hospitals and nursing homes, cruise
ships, and schools. In otherwise healthy populations,
Norwalk virus is generally mild, and self-limiting, but there
is evidence to support that it can lead to future illnesses,
and can be fatal for the elderly and otherwise
FIGURE 1
A transmission electron micrograph of Norovirus
particles. (Photo Credit: Charles D. Humphrey,
Centers for Disease Control and Prevention.
http://www.cdc.gov/media/subtopic/library/dise
ases.htm)
immunocompromised2. It is estimated by the World Health
Organization as the most common cause of death from foodborne diarrhea disease and the fourth
greatest burden in terms of disability-adjusted life years3. In the United States, noroviruses are
responsible for about 20 million cases annually, leading to an average of 70 000 hospitalizations and
up to 800 deaths every year4. The virus is particularly devastating in developing countries where it is
estimated to cause over 212 000 deaths annually3, with 70 000 of those being children5. There are
currently no vaccines or specific anti-viral therapies for Norwalk virus6, and treatment is limited to
rehydration (in order to alleviate the mal-effects of diarrhea)2. This is largely attributed to the critical
knowledge gaps about the virus.
There is very little known about the basic epidemiology of Norovirus in developing countries3.
Norovirus is predominantly spread person-to-person5, but there is a limited understanding of the
role of different age groups in virus transmission3. Foodborne transmission is only responsible for
about 15% of the transmission5. Individuals living in settings with poor sanitation and hygiene (as in
developing
FIGURE 2
(A)
(B)
(C)
(D)
This is the complete capsid structure comprised of 180 VP1
subunits. The A subunits (orange) form the pentamers at the
fivefold symmetry axes. The B (red) and C (blue) subunits
form the hexamers, whose centers are located at the
threefold axis. A C – C dimer interface forms the twofold
interface. The asymmetric unit is indicated by the triangle.
An enlarged asymmetric unit is shown. The capsid is made
from 60 of these units.
An A – B dimer with the angled interface indicated by the
arrows.
A C – C dimer with a flat interface as indicated by the
arrows.
countries) are more likely to be exposed to norovirus from multiple sources5, and poor health care
and availability to clean water makes treatment of symptoms more difficult.
The matter of vaccine development is complicated by evidence that supports the need for a
polyvalent vaccine against norovirus due to its antigenic variation3. Much like the seasonal flu
vaccine, this would require the synthesis of an updated vaccine when new pandemic strains of the
virus emerge3. The rapid mutation of Norovirus makes vaccines increasingly impractical7. An
antiviral therapeutic, however, may be useful in preventing the rapid spread of the disease in isolated
and crowded settings like cruise ships.
The human norovirus is a small (~38 nm in diameter), icosahedral non-enveloped virus. It has a
positive-sense single stranded RNA genome with three genes: ORF1, ORF2, and ORF3. ORF2
encodes the norovirus major capsid protein VP1. When ORF2 is expressed in insect cells, the VP1
that is produced self-assembles into Norwalk Virus-Like Particles (NVLPs). These NVLPs are
morphologically and antigenically similar to native norovirus1. This allows NVLPs to be studied as a
suitable substitute for human norovirus as they are easy to produce in large quantities, and are not
infectious due to the lack of genetic material. We are using the 3.4 Å resolution NVLP structure
determined via X-ray crystallography for our study8.
The Norwalk virus capsid exhibits 𝑇𝑇 = 3 icosahedral symmetry. It consists of 180 individual VP1
proteins. The VP1 protein has two domains: the shell (S) domain, and the protruding (P) domain.
The S domain makes up the spherical body of the capsid, and the P domain creates arc-like
structures that extend outwards from the capsid surface at every twofold line of symmetry. The P
domain is made of two subdomains: P1, and P29Figure 2. A flexible hinge connects the S and P
domains (shown in Figure 2). It has been shown that the S domain alone is enough to assemble the
actual capsid. A mutant VP2 with its P domain deleted was able to form a capsid (CT303) that
resembled the wild type NVLPs without the protruding domains1. This suggests that the S domain is
sufficient for the assembly of the icosahedral capsid. The P domain has been found to be critical in
several structural dynamics of the Norwalk virus capsid. It has a role in increasing the capsid
rigidity9, and in mediating host cell attachment10,11. Since the P domain is somewhat flexible, it is able
to recognize and bind to several different host antigens10,11. This adds to the pervasive infectivity of
the virus and makes it much harder to create preventative treatments.
In order to form the icosahedral capsid, the VP1 protein assumes three quasi-equivalent
conformations12, conventionally referred to as A, B, and C1 (Figure 2 A-B). The capsid assembles
from VP1 dimers. In the capsid, these dimers exist in two distinct conformations, and can be
distinguished as A-B dimers and C-C dimers. The C-C dimers exist in a relaxed conformation
analogous to the dimer configuration in free solution (Figure 2 D). In contrast, the A-B dimer
conformation is slightly strained due to their placement in the capsid (Figure 2 C). This can be seen
in the relatively linear interface between the C-C dimer, and the slightly angled interface between the
A-B dimer.
Viral entry proceeds through endocytosis by the host cell through a non-clathrin, non-caveolin
mechanism13. Once inside the host cell, the virus needs to expose its RNA to the cytoplasm where it
can be translated, and the virus can be replicated. Little is known about the uncoating process
through which this happens, but since the capsid does not have pores, the uncoating process is
thought to involve a partial or complete disassembly of the capsid. An understanding of the
mechanical strength of the various interfaces in the capsid may lend insights into requirements for
disassembly.
Atomic force microscopy (AFM) indentation experiments have been performed on NVLPs to gain
insights into the mechanical properties of the capsid. In these experiments, the viral particle is
adhered to a slide, and can either be kept in liquid solution or air for the duration of the experiment.
In imaging mode, the tip of the AFM is able to scan the surface of the NVLP, and produce a
topographical image. This allows for measurements of the size of the capsid. In indentation mode,
the AFM tip can be used to apply a controlled force on the capsid. As the indentation progresses,
the amount of force (F) and distance (Z) is recorded in what is known as an FZ curve. These
measurements can provide information about the strength and flexibility of the entire capsid. The
FZ curve typically has an initial linear slope that yields the particle’s ‘spring constant’. This is
followed by a sharp drop that correlates to the buckling or fracture of the shell14. The location of
this drop gives the critical force and indentation at which capsid failure occurs.
There have been two nanoindentation experiments done on Norwalk virus to date. The first focused
on the influence of pH on the mechanical behavior of the capsid15. In that work, when the
experiments were performed under neutral pH, the measured spring constant was 0.05 N/m. In
this study, however, the measured capsid dimensions were inconsistent with the known dimensions.
This may indicate that the particles may have been deformed in the preparation or imaging mode
leading to an inaccurate stiffness. A second study that examined the influence of the P domain on
the mechanical properties of NVLPs found the wild-type NVLP to have a spring constant of 0.30
N/m9. Their capsid dimensions were consistent with the known values, and so we were inclined to
use their values for further referencing.
Our goal in this study was to perform nanoindentation simulations on the NVLP in order to obtain
information that cannot be observed through AFM experiments. It has been demonstrated that
simulations can reasonably reproduce nanoindentation experiments. In fact, a simulation study
utilizing a coarse grained model was able to measure the spring constant of 35 capsids with good
agreement for systems with known spring constants16. Furthermore, the incredible resolution of
computational studies can give unique insights into the structure and dynamics of biological
machines. For example, in a typical (non-simulated) nanoindentation experiment, there is no way to
calculate intermolecular forces during the capsid deformation, but this can be done easily through
computational simulation of the system.
Specifically, we wish to obtain information about the relative strengths of the various proteinprotein interfaces in the Norwalk virus capsid. This may provide insights into the mechanisms of
uncoating, which could be used for development of antivirals that work by inhibiting this uncoating
process17.
Methods
Overview of Molecular Dynamics
In the Feynman Lectures on Physics, it is stated18 that “everything that living things do can be
understood in the jigglings and wigglings of atoms”. Molecular dynamics (MD) is an attempt to
study the dynamics of macromolecules, such as proteins, nucleotides, lipid bilayers, and
carbohydrates19, by examining this seemingly random movement of atoms. This is done by
integrating the classical Newtonian equations of motion for all the atoms in a system19. These
equations are integrated at very small time steps (typically 2 fs), several million times to get an
adequate amount of simulation data. At present, it is common to see total simulation trajectories that
are several hundred nanoseconds, or even microseconds long. Once the motion of the system is
known, several structural and dynamic properties can be determined for the molecule.
Each calculation is done to determine the various microstates in the system throughout the
trajectory. The microstate of each particle 𝑖𝑖 is defined as a function of its position 𝑟𝑟𝑖𝑖 and momentum
𝑝𝑝𝑖𝑖 called the Hamiltonian 𝐻𝐻. This function is a sum of the kinetic energy 𝐾𝐾 and the potential energy
𝑉𝑉 19.
𝐻𝐻(𝑟𝑟, 𝑝𝑝) = 𝐾𝐾(𝑝𝑝) + 𝑉𝑉(𝑟𝑟)
(1)
The kinetic energy takes a familiar form that depends on 𝑚𝑚𝑖𝑖 (the mass of particle 𝑖𝑖) and 𝑝𝑝𝑖𝑖𝑖𝑖 , 𝑝𝑝𝑖𝑖𝑖𝑖 , 𝑝𝑝𝑖𝑖𝑖𝑖
(the 𝑥𝑥, 𝑦𝑦, 𝑧𝑧 components of its momentum 𝑝𝑝)19.
1
2
2
2
𝐾𝐾 = ∑𝑁𝑁
𝑖𝑖=1 2𝑚𝑚 �𝑝𝑝𝑖𝑖𝑖𝑖 + 𝑝𝑝𝑖𝑖𝑖𝑖 + 𝑝𝑝𝑖𝑖𝑖𝑖 �
𝑖𝑖
(2)
The potential energy V is much more difficult to calculate and generally has to be estimated. In MD,
the potential energy function V is given as a force field, and is the sum of various interactions. There
are a variety of force fields that can be used such as AMBER20, CHARMM21, GROMOS22, and
OPLS23. Generally, the potential energy is divided into two parts in these force fields: the bonded
terms (comprised of bond, angle, dihedral, and improper interaction terms), and the non-bonded
terms (comprised of Van der Waals (vdW) and electrostatic interaction terms)19.
𝑉𝑉 = 𝑉𝑉𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 + 𝑉𝑉𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 + 𝑉𝑉𝑑𝑑𝑑𝑑ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 + 𝑉𝑉𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + 𝑉𝑉𝑣𝑣𝑣𝑣𝑣𝑣 + 𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
(3)
These terms address the different forces that may be acting on the particle at each time-point. These
interactions are calculated pairwise for all the atoms in the system. Generally, the bonded terms are
easier to compute than the non-bonded terms because they involve smaller pairs of atoms. In
contrast, a non-bonded interaction like Van der Waals can involve long range interactions that
involve all the atoms in the system. For this reason, there is generally a cutoff distance defined for
long range interactions to speed up the calculation. There are several other optimizations that are
made to simplify many of these calculations to make them easier to compute. Even after these
optimizations, simulations take a very long time to compute even on large computational clusters.
For a large system such as a viral capsid, further simplification of the system is often necessary to get
meaningful lengths of simulations.
Coarse Graining via a Gō Model
We are using a modified coarse grained Gō model for the viral capsid [insert citation for Eric May’s
paper]. A coarse grain model is necessary for the indentation simulations because the viral capsid is
too large a system to simulate at an atomistic level given the resources we have. A Gō model is
especially good for looking at protein dynamics.
The general scheme for creating a coarse-grained (CG) Gō model is that the atoms for each residue
are grouped into a single bead. Then there is a force field created based the “master contacts” in the
initial structure. A contact is defined when a pair of residues have a minimum distance of 12
angstroms or less in the initial atomistic structure. These initial distances for the contacts are favored
throughout the length of the simulation. An attractive force is felt between contact pairs. This sort
of model is well suited for our indentation simulation as it focuses the calculation on the deforming
forces on the protein, which is our parameter of interest. It can help us study the deformation of the
large viral capsid system in a much more computationally simplified way. It allows our simulations to
complete in a matter of weeks, rather than years.
As a part of this project, we also wanted to generalize our method of making Gō models of viral
capsids, as the current scripts are labor intensive and time consuming methods and require manual
intervention in making these models. The process was fully automated using original software
created by me.
We can run CHARMM Gō Model Builder to get a structure and native contacts for an expanded
asymmetric unit. This model also includes non-bonded forces between neighboring asymmetric
units, but it is challenging to map these neighboring contacts to the right neighbors for each of the
60 asymmetric units present in the capsid (Figure 3).
A generalized algorithm to identify neighboring
asymmetric units in the correct relative orientations.
Each asymmetric unit (AU) is numbered (somewhat
arbitrarily). First, we find neighbors of AU-1 by
calculating pairwise distances between atoms
composing AU-1 and those creating the rest of the
capsid. If an AU has an atom within 12 Å of AU-1,
it is considered a neighbor. Then we create a
FIGURE 3
simplified representation of the capsid by
A numbered icosahedron showing all faces
neighboring asymmetric unit 1. The problem that
needed to be addressed for the algorithm to be
successful is assigning these neighbors in the correct
position for all 60 asymmetric units (i.e. – if we were
to replace AU-1 with AU-40, which asymmetric
units would go in place of AU-2, 3, 4, 5, 16, 23, 6, 7,
and 9).
representing each AU as a center of mass (COM)
point, and an orientation basis. We transform the
COM of each neighbor of AU-1 by the orientation
basis of AU-1. This collection of vectors can now be
used to find neighbors for all the other asymmetric
units. Now, for each AU-x (where x is the AU number), we can transform the COMs of all AUs by
the orientation basis of AU-x. Then the coordinates of the neighbor vectors will point to all the
neighbors of AU-x. This approach allows us to not only find neighbors for each asymmetric unit,
but also get them in the right relative positions. (i.e. if AU-1 has two neighbors AU-2 (on the left)
and AU-3 (on the right), for any given AU, we know which neighbor is on the left and which is on
the right). Once we know how the neighbors map to all the asymmetric units, we can create the CG
model and define the native contacts for the entire capsid from transformations of the information
from the Gō Model of the expanded asymmetric unit. During this process we also converted the
CHARMM parameter files to GROMACS ITP files for defining the custom force fields to be used
for the Gō model. The total size of the capsid model was 89700 CG particles.
Modeling the Probe Tip
The probe tip was modeled as a hemisphere with a radium of ~9 nm with 550 identical 2kDa
particles arranged in a cubic lattice with a distance of 1.5 nm between the particles. Adjacent
particles in the lattice were bonded to each other, and movement is frozen in the x and y direction.
In-silico Indentation Simulations
The simulations were performed using the GROMACS simulation package version 4.5.524 using a
Langevin integrator with 1 ps-1 friction factor, 10 fs time step at a constant temperature of 270 K.
Non-bonded forces are computed for residue pairs as far as 33 Å apart, and data were saved at 100
ps intervals. The interaction between the virus and tip follows the 12-10 potential, where the
interaction becomes repulsive at distances below 15 Å. The tip is moved by applying an umbrella
potential with a spring constant of 200 kJ/(mol∙Å2) at a rate of 2 × 10-5 nm/ps. The force acting on
the virus is determined from the difference between the location of umbrella potential and center of
mass of the tip as shown in equation ,
𝐹𝐹(𝑡𝑡) = 𝑘𝑘(𝑧𝑧(𝑡𝑡) − 𝑧𝑧0 − 𝑣𝑣 × 𝑡𝑡)
(4)
where 𝑧𝑧0 is the initial tip center of mass position and 𝑣𝑣 is the umbrella velocity.
We performed a total of nine nanoindentation simulations on the CG model of an NVLP. There
were 3 trials for each of three orientations of the tip on the capsid: the tip was either on the twofold, 3-fold, or 5-fold axis of the capsid. This was done to account for orientational averaging that
occurs in experimental nanoindentations due to lack of control of viral particle adhesion to the slide.
Additionally, the residues in the bottom half of the capsid were held in fixed positions to reduce
noise in the data. Furthermore, this serves to better
replicate the experimental AFM indentations as the
capsid has to be initially stuck to a slide which leads
to deformation of the bottom half of the capsid.
FIGURE 4
(A) Schematic of experimental ADM
nanoindentation with force-indentation curves.
(B) Snapshots from a trajectory along a twofold
symmetry axis. As the trajectory progresses, and
the AFM tip pushes down, the distortion in the
capsid can be seen. Because the bottom half is
immobilized, it does not exhibit any
deformation.
Calculating the Spring Constant, Critical Force, and Critical Indentation
The spring constant, 𝑘𝑘𝑐𝑐 , relates the capsid stress and strain during the initial phase of deformation,
where the capsid displays a linear elastic character. It is determined from the initial linear region of
the force vs indentation curves.
The critical force signifies the maximum force the capsid can withstand before fracturing. The
critical indentation is the indentation at which the critical force is observed. These values are
determined from the point at which the force sharply decreases in the force vs indentation curves.
We computed the average mechanical properties by using a weighted average of the 3 orientations
based on how often they occur in the viral capsid based on there being 30 twofold, 20 threefold, and
12 fivefold sites on an icosahedron capsid. Similarly, in order to calculate a standard deviation (SD),
a distribution of the values was created by duplicating the measured values to match the ratio of
values to the symmetry probabilities. This SD served as the errors for the means.
Evaluating the Relative Strengths of Protein-Protein Interfaces
The protein-protein interface strengths were evaluated based on the amount of time it took for the
various types of interfaces to break. The interfaces were grouped by which version of the VP1
protein they were in between (A, B, or C), and were further distinguished by whether they were
between the P domains or S domains. There was one additional distinction for interactions between
the shell domains of the A and B subunits (SA-SB) in order to distinguish the dimer forming
interface, and the non-dimer interface.
A residue pair was considered to be maintaining its contact while the distance between them was less
than or equal to 1.5 times the native contact distance (as defined by the Gō model). When the
distance in a simulation frame went higher than this value, the contact was defined as broken. A
protein interface contact was defined as broken when at least 50% of the native contacts were lost
for the remaining trajectory. This analysis was performed using in-house software utilizing the
Pteros 2.0 library25.
Results and Discussion
Comparing the Simulated Mechanical Properties to the Experimental
Values
The mechanical parameters we chose to focus on were the spring constant of the capsid, the
maximum force the capsid could withstand (critical force), and the indentation depth at which the
critical force occurred (critical indentation). These values allowed us to compare our simulation to
the experimental AFM indentation by Baclayon et al.9 in order to evaluate the relevance of the
simulation to actual Norovirus capsid properties.
Our average spring constant is 𝑘𝑘 = 0.21 ± 0.05 N/m, which was in good agreement with the
experimentally reported value of 0.30 ± 0.09 N/m,9 the errors are SDs of the distribution. The SD
of the simulation data is based on duplicating the measured values such that the ratios of values
from the different symmetries are consistent with the ratio of those symmetries in the capsid. Our
spring constants ranged from 0.12 to 0.26 N/m, while the experimental range was from 0.10 to
0.65 N/m. The high stiffness seen in the experimental data were low probability events, and it is
likely that our nine simulations were not a large enough sample size to capture the full range.
Despite this, our average spring constant is in exact agreement with the experimental data, and lends
support to the validity of our model.
Our average critical force was 𝑓𝑓𝑐𝑐 = 1.6 ± 0.2 nN, and average critical indentation was 𝑧𝑧𝑐𝑐 = 6.8 ±
0.7 Å. The experimentally determined values are 𝑓𝑓𝑐𝑐 = 1.1 ± 0.9 nN and 𝑧𝑧𝑐𝑐 = 4.1 ± 3.7 Å9. Again,
the errors are SDs of the distributions, and it can
be seen that the simulated and experimental
distributions overlap. The overestimation of
these values most likely has to do with the high
force loading rate applied to the tip in the
simulation. The experimental loading rate of 50
nm/s is not feasible in MD even with a CG
model because it would take too much
computation time to get a simulation in which
significant indentation has occured. MD
nanoindentation simulations from previous
studies have ranged from 104 nm/s26 to 1010
nm/s27. Our loading rate of 107 nm/s was well
within this range. These rates may appear
extremely different, but it these are not true rates
because the dynamics in CG models are
accelerated due to a smoothed energy surface. It
is likely that the effective dynamics from the Gōlike model is at least three orders of magnitude
FIGURE 5
FZ curves along the twofold (top), threefold (middle),
and fivefold (bottom) symmetry axes. Each line
corresponds to one of three trials31.
faster than the actual simulation time28. This
would put our effective loading at 104 nm/s.
While a slower loading rate in our simulations
may improve the agreement in the calculated
mechanical properties, the overall model seems to perform well in comparison to the experimental
indentation, and is adequate for addressing the relative strengths of the protein-protein interfaces.
Relative Strengths of Protein-Protein Interfaces
We tabulated the number of different interfaces broken throughout the simulation. The relative
number of breaks accumulated by the different interfaces can be explained by the order in which
capsid assembly proceeds. In solution, the VP1 protein forms dimers before capsid formation
begins8. Accordingly, it would make sense that these are the strongest interactions. This might be
FIGURE 6
FT curves overlaid with the interface breaks during the simulation. All trials of indentation along the twofold (A-C),
threefold (D-F), and fivefold (G-I) symmetry axes are shown. All interfaces, which accumulated at least two breaks,
are shown. The colored lines represent the cumulative sum of the number of interfaces, which become broken, for
various interface types. Interfaces are considered broken once the fraction of native contacts in that interface has
dropped below 50% for a consecutive period of 5 ns31.
explained by the larger interaction area for the dimer interaction because of their P domains
interacting. Indeed, the A – B and C – C dimers generally accumulated the fewest breaks. In fact, the
C – C dimers never accumulated breaks in any of the 9 simulations. This would suggest that the C –
C dimers have a stronger interface than an A – B dimer. This might be due to the fact that the A – B
dimer interface is slightly distorted in the capsid, whereas the C – C dimer conformation in the
capsid is identical to the solution VP1 dimer. The additional strain in the A – B dimer may make it’s
interface weaker.
It is also known that a nucleation event must occur for dimer assembly into a complete capsid. Ion
mass spectrometry experiments have shown that a pentamer of A – B dimers is the likely the
assembly nucleus29. This would imply that the pentamer should be stronger than the hexamer. This
is precisely what we observe in our analysis. The SA – SA interface that forms the pentamer breaks
less often than the SB – SC interface that forms the hexamer. By and large, the SB – SC interfaces
were clearly accumulating the most breaks in 8 of the 9 simulations.
Putative target sites for B–C interface stabilization
Because our analysis indicated that the B – C interface was the weakest protein interface in the
capsid, we tried to identify binding hotspots across this interface that may serve as binding sites for
an uncoating inhibiting antiviral. We used the FTMap server to identify these sites. It uses organic
probe molecules to identify consensus binding sites on a protein surface30. We were able to run it
just on the residues that are on the B – C interface, and isolate 3 binding sites that spanned the
interface.
FIGURE 7
Identification of potential SB–SC stabilization
sites. Using FTMap three binding sites were
identified, which bound to residues in both
subunits B and C. Subunit B is drawn in blue
and subunit C is red. In the top image, the
probes at site 1 are green, at site 2 orange,
and site 3 cyan. In the lower zoom-in images,
the organic probes are pink, residues with an
atom within 3Å of a probe is drawn in as
cyan for subunit B, and purple for subunit
C31.
TABLE 1
Target residues in the three putative binding sites to stabilize the B – C interface31
Binding site
Subunit B residues
Subunit C residues
1
ALA114, PHE200, VAL201,
VAL202, ALA203
GLY121, THR188, PRO189,
ARG191
2
ALA140, GLN141, LEU144
VAL61, GLN62
3
GLN62, PHE503, VAL504
THR130
Conclusion
Our in-silico coarse grained nanoindentation simulations were able to reproduce several mechanical
properties of the Norwalk virus capsid, and give further insight into capsid dynamics that could not
have been revealed through experimental techniques alone. We were able to tabulate the relative
strengths of the protein-protein interfaces in the capsid. These strengths were consistent with the
known intermediates in capsid formation. Furthermore, we saw a clearly weak interface in the capsid
between the B – C subunits. The separation of this surface might serve as the activation event for
viral uncoating, though further experimentation is needed to confirm this hypothesis. If this is
indeed the case, a drug that can bind and strengthen this interface would be a promising antiviral
compound. Accordingly, we used FTMap to identify three sites that a potential small molecule may
attempt to bind in order to strengthen this interface.
References
1.
Bertolotti-Ciarlet, A., White, L. J., Chen, R., Prasad, B. V. V. & Estes, M. K. Structural
requirements for the assembly of Norwalk virus-like particles. J. Virol. 76, 4044–4055 (2002).
2.
Iturriza-Gómara, M. & Lopman, B. Norovirus in healthcare settings. Curr. Opin. Infect. Dis. 27,
437–43 (2014).
3.
Lopman, B. A., Steele, D., Kirkwood, C. D. & Parashar, U. D. The Vast and Varied Global
Burden of Norovirus: Prospects for Prevention and Control. PLoS Med. 13, e1001999 (2016).
4.
Hall, A. J. et al. Norovirus disease in the United States. Emerg. Infect. Dis. 19, 1198–205 (2013).
5.
Lopman, B. et al. Global burden of norovirus and prospects for vaccine development. Centers for Disease
Control and Prevention (2015).
6.
Boyd, K. J., Bansal, P., Feng, J. & May, E. R. Stability of Norwalk Virus Capsid Protein
Interfaces Evaluated by in Silico Nanoindentation. Front. Bioeng. Biotechnol. 3, 1–8 (2015).
7.
Bull, R. a, Eden, J.-S., Rawlinson, W. D. & White, P. a. Rapid evolution of pandemic
noroviruses of the GII.4 lineage. PLoS Pathog. 6, e1000831 (2010).
8.
Prasad, B. V. V. X-ray Crystallographic Structure of the Norwalk Virus Capsid. Science (80-. ).
286, 287–290 (1999).
9.
Baclayon, M. et al. Prestress strengthens the shell of norwalk virus nanoparticles. Nano Lett.
11, 4865–4869 (2011).
10.
Katpally, U., Wobus, C. E., Dryden, K., Virgin, H. W. & Smith, T. J. Structure of antibodyneutralized murine norovirus and unexpected differences from viruslike particles. J. Virol. 82,
2079–88 (2008).
11.
Katpally, U. et al. High-resolution cryo-electron microscopy structures of murine norovirus 1
and rabbit hemorrhagic disease virus reveal marked flexibility in the receptor binding
domains. J. Virol. 84, 5836–41 (2010).
12.
CASPAR, D. L. & KLUG, A. Physical principles in the construction of regular viruses. Cold
Spring Harb. Symp. Quant. Biol. 27, 1–24 (1962).
13.
Perry, J. W. & Wobus, C. E. Endocytosis of murine norovirus 1 into murine macrophages is
dependent on dynamin II and cholesterol. J. Virol. 84, 6163–76 (2010).
14.
Roos, W. H., Bruinsma, R. & Wuite, G. J. L. Physical virology. Nat. Phys. 6, 733–743 (2010).
15.
Cuellar, J. L., Meinhoevel, F., Hoehne, M. & Donath, E. Size and mechanical stability of
norovirus capsids depend on pH: a nanoindentation study. J. Gen. Virol. 91, 2449–56 (2010).
16.
Cieplak, M. & Robbins, M. O. Nanoindentation of 35 virus capsids in a molecular model:
relating mechanical properties to structure. PLoS One 8, e63640 (2013).
17.
Mckinlay, M. A., Pevear, D. C., York, N. & Rossmann, M. G. Treatment of the picornavirus
common cold by inhibitors or viral uncoating and attachment. Annu. Rev. Microbiol. 46, 635–
54 (1992).
18.
Karplus, M. & McCammon, J. A. Molecular dynamics simulations of biomolecules. Nat.
Struct. Biol. 9, 646–652 (2002).
19.
Wang, Y. & Mccammon, J. A. Computational Modeling of Biological Systems. (2012).
doi:10.1007/978-1-4614-2146-7
20.
Weiner, P. K. & Kollman, P. a. AMBER: Assisted model building with energy refinement. A
general program for modeling molecules and their interactions. J. Comput. Chem. 2, 287–303
(1981).
21.
Vanommeslaeghe, K. et al. CHARMM general force field: A force field for drug-like
molecules compatible with the CHARMM all-atom additive biological force fields. J. Comput.
Chem. 31, 671–690 (2010).
22.
Schmid, N. et al. Definition and testing of the GROMOS force-field versions 54A7 and 54B7.
Eur. Biophys. J. 40, 843–856 (2011).
23.
Jorgensen, W. L. & Tirado-Rives, J. The OPLS [optimized potentials for liquid simulations]
potential functions for proteins, energy minimizations for crystals of cyclic peptides and
crambin. J. Am. Chem. Soc. 110, 1657–1666 (1988).
24.
Van Der Spoel, D. et al. GROMACS: fast, flexible, and free. J. Comput. Chem. 26, 1701–18
(2005).
25.
Yesylevskyy, S. O. Pteros: Fast and easy to use open-source C++ library for molecular
analysis. J. Comput. Chem. 33, 1632–1636 (2012).
26.
Kononova, O. et al. Structural transitions and energy landscape for Cowpea Chlorotic Mottle
Virus capsid mechanics from nanomanipulation in vitro and in silico. Biophys. J. 105, 1893–
903 (2013).
27.
Zink, M. & Grubmüller, H. Mechanical properties of the icosahedral shell of southern bean
mosaic virus: a molecular dynamics study. Biophys. J. 96, 1350–63 (2009).
28.
Buck, P. M. & Bystroff, C. Constraining local structure can speed up folding by promoting
structural polarization of the folding pathway. Protein Sci. 20, 959–69 (2011).
29.
Uetrecht, C., Barbu, I. M., Shoemaker, G. K., van Duijn, E. & Heck, A. J. R. Interrogating
viral capsid assembly with ion mobility-mass spectrometry. Nat. Chem. 3, 126–32 (2011).
30.
Brenke, R. et al. Fragment-based identification of druggable ‘hot spots’ of proteins using
Fourier domain correlation techniques. Bioinformatics 25, 621–7 (2009).
31.
Boyd, K. J., Bansal, P., Feng, J. & May, E. R. Stability of Norwalk Virus Capsid Protein
Interfaces Evaluated by in Silico Nanoindentation. Front. Bioeng. Biotechnol. 3, (2015).